Flint 2 (MT6000) 2.5Gbe issues

Technical Support for Routers
42

Sending:

[ ID] Interval           Transfer     Bitrate
[  7]   0.00-10.00  sec  2.64 GBytes  2.27 Gbits/sec                  sender
[  7]   0.00-10.01  sec  2.64 GBytes  2.27 Gbits/sec                  receiver

Receiving:

[ ID] Interval           Transfer     Bitrate         Retr
[  7]   0.00-10.01  sec  2.25 GBytes  1.93 Gbits/sec    0             sender
[  7]   0.00-10.01  sec  2.22 GBytes  1.91 Gbits/sec                  receiver

So receiving is also worse than sending, but not as bad as yours. Maybe Realtek on both ends makes it worse?

Given that I get symmetric speeds on the Mac Studio, I am not surprised if it's Realtek shenanigans (as usual).

1 Like
43

Steep learning curve but I think the MikroTik RB-5009 can, although it only has 1 10G port so you’d also need a switch with vlan support

44

That would work, but I am not sure if the RB5009 can do 4Gbit PPPoE + VLAN tagging with or without hardware offloading.

45

I’ve asked him to confirm but I think that’s what a friend of mine was using on a 4gbit KPN service.

Edit: he confirmed he’s using an rb5009

1 Like
46

Nice! Running OpenWrt looks quite involved though (not a fan of RouterOS).

47

Can we have any response or acknowledgment from gl-inet? Is this going to be investigated?

2 Likes
48

Hardware acceleration is generally between NAT, such as WAN-LAN, like server is in the WAN side, and the client is in the LAN.
Non-NAT, like LAN-LAN will not use hardware acceleration.

For the bidirectional refers to the WAN-LAN that uses hardware acceleration.

49

Alright but what about LAN / LAN? Say I have two 2.5Gbe clients with the WAN/LAN ports set up as LAN. Check my results here

50

That's not the experience I had doing intervlan routing on the flint2, there was a noticeable difference between HW acceleration on and off.

LAN - LAN traffic may not even touch the router at all depending on the network setup and in a standard setup it's likely to only need to switch chip (if the devices are using the inbuilt ports).

It's only if you have setup multiple LANs/Vlans any routing will be involved for LAN > LAN traffic.

51

The two 2.5G ports in the hardware layer are not under the same switch chip, but in LAN-LAN, so they will touch the router CPU.
VLANs/Multi-LANs, same the above I think.

52

That means the LAN-LAN, will not use hardware acceleration, the traffic via or say touch the router.

53

I was assuming the use of the 1G LAN ports in a "typical" setup.

Most people won't be doing inter vlan routing, and as such the LAN > LAN traffic is either not going to touch the router at all or if it is then it's likely to only touch the switch.

Yes there are crazy people like me who hang vlans off the 2.5G port or make the other 2.5G WAN port a LAN port in Luci but I'd imagine that's not a common config.

Single stream Iperf between VLAN's was definitely faster with HW acceleration on vs off, but I suspect enabling HW accell also enables the SW flow offloading and it's possible that is enough to speed it up.

I'd have to watch CPU load more closely to see what it's doing.

Edit:

Yep there's a noticeable difference between HW acceleration on, SW acceleration and OFF.

I was testing from an WSL2 container so that potentially causes a bottleneck

[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-30.01 sec 205 MBytes 57.3 Mbits/sec 280 sender
[ 5] 0.00-28.94 sec 204 MBytes 59.0 Mbits/sec receiver
[ 7] 0.00-30.01 sec 174 MBytes 48.6 Mbits/sec 200 sender
[ 7] 0.00-28.94 sec 173 MBytes 50.1 Mbits/sec receiver
[ 9] 0.00-30.01 sec 204 MBytes 57.0 Mbits/sec 144 sender
[ 9] 0.00-28.94 sec 202 MBytes 58.7 Mbits/sec receiver
[ 11] 0.00-30.01 sec 197 MBytes 55.0 Mbits/sec 251 sender
[ 11] 0.00-28.94 sec 195 MBytes 56.6 Mbits/sec receiver
[ 13] 0.00-30.01 sec 158 MBytes 44.2 Mbits/sec 268 sender
[ 13] 0.00-28.94 sec 157 MBytes 45.6 Mbits/sec receiver
[ 15] 0.00-30.01 sec 285 MBytes 79.6 Mbits/sec 427 sender
[ 15] 0.00-28.94 sec 282 MBytes 81.6 Mbits/sec receiver
[ 17] 0.00-30.01 sec 151 MBytes 42.2 Mbits/sec 232 sender
[ 17] 0.00-28.94 sec 150 MBytes 43.5 Mbits/sec receiver
[ 19] 0.00-30.01 sec 189 MBytes 52.8 Mbits/sec 224 sender
[ 19] 0.00-28.94 sec 186 MBytes 54.1 Mbits/sec receiver
[ 21] 0.00-30.01 sec 149 MBytes 41.6 Mbits/sec 181 sender
[ 21] 0.00-28.94 sec 148 MBytes 42.9 Mbits/sec receiver
[ 23] 0.00-30.01 sec 182 MBytes 51.0 Mbits/sec 226 sender
[ 23] 0.00-28.94 sec 181 MBytes 52.5 Mbits/sec receiver
[SUM] 0.00-30.01 sec 1.85 GBytes 529 Mbits/sec 2433 sender
[SUM] 0.00-28.94 sec 1.83 GBytes 545 Mbits/sec receiver

Router CPU was 25% SW interrupts

HW acceleration on

> 
> [ ID] Interval           Transfer     Bitrate         Retr
> [  5]   0.00-30.00  sec  1.06 GBytes   303 Mbits/sec  730             sender
> [  5]   0.00-28.95  sec  1.06 GBytes   313 Mbits/sec                  receiver
> [  7]   0.00-30.00  sec   895 MBytes   250 Mbits/sec  1045             sender
> [  7]   0.00-28.95  sec   893 MBytes   259 Mbits/sec                  receiver
> [  9]   0.00-30.00  sec   430 MBytes   120 Mbits/sec  539             sender
> [  9]   0.00-28.95  sec   426 MBytes   123 Mbits/sec                  receiver
> [ 11]   0.00-30.00  sec  1.02 GBytes   291 Mbits/sec  1046             sender
> [ 11]   0.00-28.95  sec  1.02 GBytes   301 Mbits/sec                  receiver
> [ 13]   0.00-30.00  sec   481 MBytes   135 Mbits/sec  670             sender
> [ 13]   0.00-28.95  sec   478 MBytes   139 Mbits/sec                  receiver
> [ 15]   0.00-30.00  sec   631 MBytes   176 Mbits/sec  779             sender
> [ 15]   0.00-28.95  sec   628 MBytes   182 Mbits/sec                  receiver
> [ 17]   0.00-30.00  sec   840 MBytes   235 Mbits/sec  670             sender
> [ 17]   0.00-28.95  sec   837 MBytes   242 Mbits/sec                  receiver
> [ 19]   0.00-30.00  sec   717 MBytes   200 Mbits/sec  840             sender
> [ 19]   0.00-28.95  sec   714 MBytes   207 Mbits/sec                  receiver
> [ 21]   0.00-30.00  sec   680 MBytes   190 Mbits/sec  605             sender
> [ 21]   0.00-28.95  sec   677 MBytes   196 Mbits/sec                  receiver
> [ 23]   0.00-30.00  sec   598 MBytes   167 Mbits/sec  897             sender
> [ 23]   0.00-28.95  sec   594 MBytes   172 Mbits/sec                  receiver
> [SUM]   0.00-30.00  sec  7.22 GBytes  2.07 Gbits/sec  7821             sender
> [SUM]   0.00-28.95  sec  7.20 GBytes  2.14 Gbits/sec                  receiver

Not seeing CPU usage on the router.

54

Hello
So you’re saying that when I did my iPerf tests between my nas in the lan (connected to the wan-lan port) and the router itself , it’s the same as if I did the test between wan and lan ??
I’ M not sure I understood correctly.

55

Seems the same if the iperf3 server in NAS and NAS in the WAN, Router SSH iperf3 client.

You can observe whether the traffic of the WAN (eth1) and the br-lan (or eth0) increases at the same time, to determine whether the traffic path is WAN-LAN.